00100	.SEC EVALUATION AND IMPROVEMENTS
00310	
00320	.SS Evaluation
00400		 The  primary  aim  in
00500	developing this  model  was  to  explore  and test a theory having
00600	explanatory authenticity. To satisfy this aim the model  must  meet
00700	norms  of  internal  consistency and norms of external correspondence
00800	with observation. A secondary aim would involve pragmatic norms  of
00900	application.  These  aims  are  not  unrelated  but the first is more
01000	fundamental  since  useful  applications  require  some   degree   of
01100	consistency and authenticity.
01200		As emphasized in Chapter 2 a model in the form of an algorithm consists of a structure of
01300	functions or procedures whose inner workings are sufficient to reproduce the outward
01400	symbolic behavior under consideration. The theory embodied in the model is
01500	revealed by the set of statements which makes intelligible the connections
01501	betweeen input and output by describing how the structure
01600	reacts under various circumstances.
01700		Theories have many functions. They can be summarized as follows
01800	[from Bunge?]
01850	.V
01900		(1)To systematize knowledge.
02000		(2)To explain facts by showing how they are the entailed
02100	              consequences of the systematizing hypotheses.
02200		(3)To increase knowledge by deriving new facts.
02300		(4)To enhance the testability of hypotheses by connecting  
02400		      them to observations.  
02500		(5)To guide research by:
02600	              (a)posing fruitful problems
02700	              (b)suggesting new data to gather
02800	              (c)opening new lines of investigation
02900		(6)To map a portion of reality.
02950	.END
03000		It is a tall order for a theory to fulfill all of these
03100	functions. In undeveloped fields we should be happy with even one  of
03200	them.   Models   can  be  assigned  these  functions  when  they  are
03300	theoretical in type. Our model was intended
03400	primarily to serve functions (2) and (4), testable explanation.
03500		What constitutes a satisfactory explanation has been  treated
03600	in  2.1.  The  "fit"  or  correspondence  with phenomena as
03700	indicated by measurements and empirical tests indicate the degree of faithfuleness
03800	of the reproduction, the authenticity of the model.
03900		Decision procedures for a consensus  acceptability  of a model
04000	sometimes depend not so much on truth, an elusive state, as
04100	on whether a majority of the relevant expert community  believes  the
04200	theory  or  model to approximate truth to some unknown and unknowable
04300	degree and be better than available alternatives. Truth or  falsity  cannot  be
04400	proven  with certainty but their presence can be assayed by some sort
04500	of critical assesment and deliberation. A theory or model  may  bring
04600	cognitive or pragmatic comfort, not because it is TRUE but because
04700	it represents an improvement  over its  contending  rivals.  Cognitive
04800	comfort  is  a  type  of  intellectual  satisfaction  while pragmatic
04900	comfort accrues from applications to problems in order to make things
05000	work the way humans want them to work optimally in practical contexts 
05100	of action.
05110		Although I believe that faithful reproduction is the major test  for
05200	the adequacy of symbolic models, it would be a bonus if our model could  satisfy  
05300	function  (3)  ,  making  possible  new
05400	knowledge  through  prediction.  It   would   give   clinicians   and
05500	investigators  something to look for. This novelty could arise in two
05600	ways. First, the model might demonstrate a property  of  the  paranoid
05700	mode  hitherto  unobserved clinically.   In principle this could come
05800	about because the I/O behavior of the model is  a  consequence  of  a
05900	large   number  of  interacting  hypotheses  and  assumptions  chosen
06000	initially  to explain frequently observed phenomena.  When  the
06100	elements of such a complex conjunction interact with highly variable inputs they
06200	generate consequences in addition to  those  they  were  designed  to
06300	explain.  Whether  any  of  these  consequences  are  significant  or
06400	characteristic of the paranoid mode  remains a subject for far-in-the-future
06500	study.
06600		Another bonus would lie in the behavior of the model
06700	in some new situation. Since it is designed to simulate communicative
06800	behavior in an interview situation, the `new' circumstance would have to
06900	involve some new type of linguistic interaction which the model is capable
07000	of responding to. From its behavior one might then predict how paranoid
07100	patients would behave under similar circumstances. The requisite
07200	empirical tests and measures would show the degree of correspondence
07300	between patient and model behaviors.
07400		This possibility is of importance  in  considering  therapies
07500	for  patients  tangled in the quandaries of the paranoid mode.  Since
07600	the  model  operates  at  a  symbol processing  level  using  natural
07700	language, it is at this level at which linguistic and semantic skills
07800	of clinicians can be applied. Language-based or semantic  techniques  do  not  seem
07900	very effective in the psychoses
08000	but they are useful in states of lesser severity. A wide range of
08100	new semantic techniques, including extremes, could be tried first on
08200	the model without hurting patients through blind experimentation.
08300		While our group has used the model to explore a theory and to
08400	study psychiatric judgements, its potential use as a training device
08500	has not escaped our attention. Medical students and psychiatric residents
08600	need `disposable' patients to practice on without jeopardy.
08700	The paranoid model can print out a trace of its inner states 
08800	during and after an interview. Whether the optimal goal of interviewing
08900	gathering relevant information without upsetting the patient, has been
09000	achieved, thus can be estimated. A beginning interviewer can practice
09100	in private or with a supervisor present. Many interviewers have reported
09200	that the model has a definite effect on them. The student can get the
09300	feel of the paranoid mode long before he interviews an actual patient.
09400	The effect of various interviewing styles can be studied and compared.
09500	.SS Extra-evidential Support
09600		Besides the function of evidence in evaluating models, there is
09700	is the role of extra-evidential support. This support derives from
09800	plausibility arguments, comparisons with competing models or theories,
09900	and coherence with other domain theories.
10000		A theoretical model is evaluated relative to rival explanations.
10100	The model I have described stands as a candidate for a consensus  explanation
10200	of paranoid processes. It has greater explanatory power than ,say, the  
10210	"homosexual-conflict" theory because it covers a more comphrehensive 
10220	range of facts. The expert forum will decide its ultimate status.
10300	(More here on plausibility, competitors, and other domain theories?)
10400	(more comprehensive= covers more facts→ greater explanatory power, but	
10500	accuracy more important than range.)
10600	
10700	.SS Improvements
10800	
10900		As the validation studies indicated further work must
11000	be carried out to improve the model's linguistic limitations. At
11100	the moment several co-workers are collaborating in this effort- Horace
11200	Enea, Charles Rieger, and William Faught. So many special cases arise
11300	in unrestricted natural-language dialogues that a large number of
11400	rules must be written to cover them.
11500		Future versions of the model will have a greatly expanded
11600	data-base or "memory" which will allow it to answer a  many more
11700	questions than the version described herein. The memory must be
11800	interfaced with the language-analyzer in such a way that when some
11900	input expression is not well understood, the memory can still attempt
12000	an appropriate answer. This capability is especially needed when the
12100	interviewer refers back to something already mentioned in the interview.
12200		Another improvement involves the ability of the model to
12300	build up a model of the interviewer as the interview proceeds. Information
12400	from this submodel can then be used in contending with the interviewer
12500	who is believed to be malevolent. Instead of making rather primitive
12600	general remarks abot the interviewer the model will be able to be
12700	much more specific in its comments.
12800		Finally the theory must be extended so that the model is
12900	dynamic over time. The version described here changes only over the
13000	course of a single interview. To explore how changes can be brought
13100	about through external input, the model must have capabilities for
13200	self-modification over longer periods of time in which it interacts
13300	with a number of interviewers. A further use of  more dynamic models
13400	will be to explore the ontogenesis of paranoid thought, that is, how
13500	a system grows to be the way it is through long periods of socialization.